Unveiling the veil: Identifying potential shell firms using machine learning approaches

Abstract

China's approval-based initial public offering (IPO) system has fostered a shadow market of undisclosed potential shell firms, which play a crucial role in enabling reverse mergers (RMs) that bypass IPO regulatory scrutiny. Using machine learning (ML) techniques and firm-level data from 2011 to 2021, we identify these hidden shell firms and examine their characteristics. We find that shell firms are typically overvalued and exhibit weaker sensitivity to market-wide movements. Compared with traditional logistic models, the ML model demonstrates superior predictive and explanatory power in distinguishing shell firms from regular firms. Benefit–cost analyses further show that investors, auditors, and regulators can derive meaningful benefits from the model while incurring minimal costs. We contribute to the literature by applying ML to uncover hidden shell firms and by highlighting market inefficiencies arising from IPO entry restrictions.